Nvidia's up to $500B AI data-center plan hinges on backstopping aging GPUs

Nvidia announced this week that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR were willing to commit up to $500 billion to build AI data centers. The headline figure drew attention, but the more significant move is Nvidia’s attempt to create a secondary market for aging GPUs: to win over the financial firms, Nvidia agreed to guarantee, with its own money, that chips used as collateral in these deals will retain their value.

Specifically, Nvidia is promising that if the GPUs backing a loan don’t hold their expected value — for instance, if a data center owner defaults and a lender has to liquidate but the chips can’t fetch the price the books assumed — Nvidia will cover up to 25% of the shortfall. That creates what financiers call “wrong way” risk: Nvidia’s obligations grow precisely as demand for its chips weakens, which is also when its own revenue would likely be under pressure.

The plan has drawn comparisons to Lucent Technologies, the telecom equipment maker that lent its customers money to buy its gear and then crashed alongside them in the dotcom bust. Nvidia has already committed billions to buyers of its chips, including frontier labs OpenAI and Anthropic and neoclouds like CoreWeave, Nebius, Firmus, and Lambda, and Bloomberg has calculated it’s working on another $750 billion worth of circular deals this summer. “Is this circular financing?” Nvidia CEO Jensen Huang wrote on X about the new scheme. “This initiative is designed to address that concern. We are bringing independent, long-term institutional capital into the AI infrastructure market.”

Unlike Lucent, Nvidia is getting outside investors to shoulder most of the capital and risk, merely agreeing to protect a portion of its chips’ future value. If the bet pays off, it gives Nvidia a new funding channel for AI data centers just as older methods — debt at Oracle, new equity at Google, heavy cash burn at Meta — show strain. Microsoft CEO Satya Nadella recently recommended a book about the railroad-era financial engineering that crashed the U.S. economy in 1873 during an earnings call, an implicit nod to the risk.

Why it matters

Nvidia is betting that a broad ecosystem of buyers will keep demand for its GPUs — new and aging — deep enough to make its up to 25% guarantee cheap to honor; if AI demand growth merely slows rather than keeps outpacing supply, that bet gets expensive fast.